Magellan Final Report - Office of Science - U.S. Department of Energy
Magellan Final Report - Office of Science - U.S. Department of Energy
Magellan Final Report - Office of Science - U.S. Department of Energy
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Magellan</strong> <strong>Final</strong> <strong>Report</strong><br />
Productivity. One advantage <strong>of</strong> cloud computing can be increased productivity. For users who need the<br />
added flexibility <strong>of</strong>fered by the cloud computing model, additional costs may be more than <strong>of</strong>fset by the increased<br />
flexibility. Furthermore, in some cases the potential for more immediate access to compute resources<br />
could directly translate into cost savings. Section 12.5 discusses examples where this could occur. Potential<br />
increases in productivity should be weighed against any up-front investments required to re-engineer s<strong>of</strong>tware<br />
or re-train staff.<br />
Acquisition Costs. It is generally assumed that cloud providers pay significantly lower costs for hardware<br />
than typical IT customers. This assumption is reasonable given the purchasing power a cloud provider can<br />
exercise. Since cloud providers do not publish their acquisition costs, it is difficult to make direct comparisons.<br />
However, DOE centers use competitive procurements or close partnerships to procure their systems. These<br />
procurements are typically very large and command substantial discounts from both the integrator and<br />
component providers. These discounts may not be quite as high as for warehouse scale datacenters, but<br />
likely approach them. Furthermore, the purchasing contracts typically include acceptance criteria tied to<br />
application performance which significantly reduces risks.<br />
12.4 Historical Trends in Pricing<br />
One argument <strong>of</strong>ten made in favor <strong>of</strong> commercial clouds is that it enables customers to automatically<br />
reap the cost benefits <strong>of</strong> technology improvements. HPC customers have come to count on the “Moore’s<br />
Law” improvements in computing capability which typically result in a doubling in capability per dollar<br />
approximately every two years. As a result, DOE centers have historically delivered average improvements<br />
in computing capability <strong>of</strong> 40%-80% per year with relatively flat budgets. However, this level <strong>of</strong> decreasing<br />
costs for compute performance is not currently observed in the commercial cloud space. For example, the cost<br />
<strong>of</strong> a typical compute <strong>of</strong>fering in Amazon (m1.small) has fallen only 18% in the last five years with little or no<br />
change in capability. This translates into a compound annual improvement <strong>of</strong> roughly 4%. Meanwhile, the<br />
number <strong>of</strong> cores available in a typical server has increased by a factor <strong>of</strong> 6x to 12x over the same period with<br />
only modest increases in costs. While there are examples <strong>of</strong> cost reductions in commercial cloud services,<br />
such as free in-bound data transfers or lower storage cost for higher tiers, the overall pricing trends, especially<br />
for computation, still show only modest decreases. The new systems coming online at the ALCF provide an<br />
example <strong>of</strong> how DOE centers delivere increased computing capability with relatively flat budgets. The three<br />
new IBM Blue Gene/Q systems at the ALCF will provide an estimated total HPL Peak <strong>of</strong> 8.4PF (based<br />
on prototype Blue Gene/Q systems currently listed on the November 2011 Top500 list) at an anticipated<br />
annual budget <strong>of</strong> $62M/year. This results in a Cost per TF to operate for 1 year <strong>of</strong> $7.4K, which is 13-14x<br />
better than Amazon’s calculated cost (Section 12.2.4), and is an 11x improvement over the current ALCF<br />
resource, Intrepid, which is around four years old.<br />
12.5 Cases where Private and Commercial Clouds may be Cost<br />
Effective<br />
There are some cases where moving a workload to a private or public cloud <strong>of</strong>fering can be cost effective.<br />
We will briefly describe a few <strong>of</strong> these cases.<br />
Unknown Demand. In the case <strong>of</strong> a new project or a new application area where the potential demand is<br />
still poorly understood, it can be cost effective to use a commercial cloud while the demand is quantified and<br />
understood. Once the level has been established, in-house resources can deployed with greater confidence.<br />
Similarly, if a new project’s prospects are highly uncertain, clouds can be a cost effective option while the<br />
long-term fate is determined. In both cases, savings are achieved by avoiding investments in unneeded ca-<br />
125